I've spent a lot of time trying to get LLM to generate things in a specific way, the biggest take away I have is, if you tell it "don't do xyz" it will always have in the back of its mind "do xyz" and any chance it gets it will take to "do xyz"
When working on art projects, my trick is to specifically give all feedback constructively, carefully avoiding framing things in terms of the inverse or parts to remove.
Funnily enough, that is true also for giving instructions to kids. And also why kid's media is so frustrating. So many shows and books focus first on the maladjusted behavior, with the character learning not to the-bad-thing at the very end.
Don't instruct kids, nor LLMs via negativa.
If I as a human deploy code, it is not certain that it necessarily works - just like with LLMs. The extent is different however.
Cool thing is that, since LLMs are comparatively cheap, I can afford to run the same process a few times, to get a sense of confidence of the response.
In our latest project, the client expressed that our AI aided process was 11 times faster, and much more accurate than their previous process.